(24) 



represent the individuals, are the average spine numbers 

 of their five (in a few cases, six) progeny. All indivi- 

 duals which did not produce at least five progeny ?:ave 

 been omitted from the table. 



On December 14, after four selections had been made in 

 both high and low groups, selection was discontinued, and 

 the final iiidividuals 5alal and 5c2a2 and their progeny 

 v/ei'e allowed to multiply freely. Because of cold weather, 

 the fission rate was very low, and when about fifty indi- 

 viduals had been produced in each group the experiment was 

 ended on February 3, 1916. 



The numbers concerned in this experiment are not large, 

 but the difference between the means of the two final groups 

 ( 1.46 spines ) is go great in proportion to its probable 

 error ( .035 ) that there can be no doubt that in this case 

 selection has isolated, within a single clone, two strains 

 with decidedly different average spine num.bers . In Figure 

 6 are shoy/n a fev/ characteristic members of the two groups, 

 from camera outlines. 



In this experiment there is one possibility of error. 

 When one is selecting individuals instead of masses of in- 

 dividuals, a single line instead of a number of linos, an 

 excellent opportunity is offered for the multiplication of 

 any mutation that may occur. Recognizing this fact, and 

 being now quite convinced in m.y own mind that selection 

 within the clone was having an effect, I turned to mass 

 selection once more. In mass selection, with many paral- 

 lel lines, there is no chance for the extreme multiplica- 

 tion of a mutation, and if one does occur it changes the 



